[trishorts]

This ports pyMzLib #87 to mzLibR. It documents every result change that mzLib 1.0.595 brought with the 0.5.0 bridge, and it drops the package's own PRIDE timeout.

What changes

• pride_list_files(), pride_list_ftp_files() and pride_search() default to timeout = NULL, not 300. mzLib already bounds each request: three retries after 5, 20 and 60 s, with each attempt up to 100 s (#1412). A 300 s limit of the package's own therefore turned a hanging PRIDE outage into a timeout instead of mzlib_service_unavailable. Pass timeout = 300 for the old limit.
• NEWS.md now has every result change pyMzLib's upgrading guide lists for the 0.5.0 bridge, in R terms. Each one was re-checked live from R against the v0.5.0 bridge:
  • ProForma input refusals (#1462);
  • SUMO names on D now fail and on K convert (#1432);
  • TMTpro on S/T/Y/H now fails, while on K gives +304.2071 (#1419);
  • Lactylation is 72.021129 Da with UNIMOD:2114 (#1417);
  • the entrapment flags and entrapment_fraction (#1371);
  • FlashLFQ normalization and gapped designs (#1422, #1424);
  • |P proteases at sequence ends (#1320);
  • stats_fit() below 3 features (#1398);
  • PRIDE retries and .partial resumption (#1412);
  • the download checks that don't reach pride_download() (#1427, #1449);
  • contrasts (#1452).
• Help and vignettes:
  • ?stats_fit says prior$scale is NA with fewer than 3 features, and that bridges before pyMzLib 0.5.1 report the first feature's variance there. Today's 0.5.0 bridge does that: a 2-feature fit shows 0.0251.
  • ?pride_download covers the resumable .partial pairs.
  • ?mzlib_error names the two EBI failures that still arrive as a bridge error after mzLib's retries: an empty FTP listing and an EBI FTP 403.
  • The PRIDE vignette gains "Retries, timeouts and outages" and "Interrupted and suspect downloads".
  • The stats help and vignette say mzLib has contrasts but stats_fit() does not expose them yet.
• Generated files: specs re-synced at bridge f85fef7, 49 help pages rebuilt, and docs/name-parity.md regenerated (header only).

Tests

The new "PRIDE's metadata calls add no timeout of their own" test fails on the old code, where the stand-in runner sees 300 for all three functions. A second test checks that an explicit timeout still reaches the bridge.

Verified locally with R 4.6.1:

• offline suite: 429 passed, 0 failed, 40 skipped;
• with the v0.5.0 bridge staged: 457 passed, 0 failed. The 12 skips need pyMzLib's fixture folders.
• sync-specs, stage-replay, render-cite, docs-lint, build-man and name-parity are all current;
• R CMD check --as-cran --no-manual --ignore-vignettes: OK. Pandoc is not installed locally, so CI builds the vignettes.

One known gap, recorded upstream as PRIDE-B3: until mzLib marks its exhausted transient failures, an empty FTP listing or an EBI FTP 403 arrives as mzlib_bridge_error. The live canaries skip only on mzlib_service_unavailable, so during such an outage they fail instead of skipping.

🤖 Generated with Claude Code